Comparative analysis of enacts gridded Rainfall Dataset with observations at twelve synoptic stations in Rwanda

Abstract

In this study, a comparison of ENACTS gridded rainfall data with 12 synoptic stations rainfall measurement namely Kanombe, Gisenyi, Rubengera, Kamembe, Gikongoro, Rubona, Byimana, Ruhengeri, Byumba, Nyagatare, Kawangire and Kibungo Kazo was carried out on monthly basis using statistical analysis of Bias, RMSE and person correlation coefficient. The results have indicated positive or negative biases changing from month to month and from station to station as shown in Figure 3. A positive bias indicated overestimation while a negative showed underestimation of station rainfall.For example, the ENACTS underestimated rain gauge measurement at Byimana station for the months of January and August, Rubengera for the months of June, November and December, Byumba for February, Kamembe for March, Kibungo Kazo for April, Gikongoro for May, Rubona for July, Kawangire for September and Kanombe in October while in March, ENACTS overestimated rain gauge measurement at Rubona, Rubengera and Byumba stations. Figure 4 and Figure 5 show respectively the results from RMSE and correlation analysis and have shown a good performance of ENACTS rainfall data expect the month of April where there was a big magnitude of error (RMSE >1) and the month of February that was poorly correlated. In this study, ENACTS gridded rainfall has indicated to be a good representative of rainfall for the study stations despite the few differences that were shown by the analysis of bias, RMSE and correlation which need to be corrected if ENACTS is used inter-changeably with station rainfall data.